collapse issues when training

[gengyuanmax]

Hi,

I try to train my model without superpixel planar regularization. But it's likely that the training collapses and the output is all-zero. Have you faced and managed to solve this problem?

Thanks in advance.

[niujinshuchong]

@GengyuanMax

Sorry for the late reply. Could you describe your training setting (e.g. which datasets) in more details?

[ShuweiShao]

Hi, I encounter the same problem. When I set the weights of lambda_planar_reg term as 0.05, the framework can derive plausible results. However, when I simply modify 0.05 to 0, the training collapses and the depth predictions seem to be all-zero.

[ShuweiShao]

It should be noted that, the dataset I used is of endoscopic scenes. In theory, the framework should work better when remove the planar regularization